leveraging data analytics for customer support efficiencymployee loyalty analysis based on employee...
TRANSCRIPT
Leveraging Data Analytics for Customer Support Efficiency
Kelly HoopesSenior Consultant
Service Strategies
October 6, 2016
OverviewDiscussion Topics
• Types of Data
• Structured vs. Unstructured Reporting
• Tips for Working with Unstructured Data
• Unstructured Data Analytics Readiness
• Use Case: Support Issue / Brief Description
• Other Practical Examples & Ideas
Types of Data: Structured Data
• Stored in relational SQL database• Easily mapped into pre-designed fields• Simplest way to manage information• Represents only 5 to 10% of all informatics
data
5-10%
"A DBA walks into a restaurant looking for a job as a busboy. He gets hired based in his experience with cleaning tables and working with servers.”
Types of Data: Semi Structured Data"A DBA walks into a NoSQL bar, but turns and leaves because he couldn't find a table"
• Doesn’t reside in a relational database • Some organizational properties and tagging
makes it easier to analyze• Semi structure exists to ease space, clarity or
to compute
• Examples: NoSQL, CSV, JSON,XML 5-10%
Types of Data: Unstructured Data”I hate the term ‘Big Data’. It’s all just data. Someday we will call it just data, but for now it’s ‘Big’.”
• Often includes text and multimedia content• Mostly referred to as “Big Data”• Unstructured data is everywhere
• Examples: Satellite Images, Sonar Data, Photos & Videos, Social Media, Web Content, Text Messages, Enterprise Data (survey results, logs, emails, customer communication) 80%+
Reporting: Structured vs. Unstructured
STRUCTURED
• Mostly clean• Easy to report• Industry prescriptive reporting
& protocols
• Very banal (so lacking in originality as to be obvious and boring)
UNSTRUCTURED
• Messy, not clean• Difficult & relatively new discipline• Non-prescriptive & no protocols
• Creative & fascinating• Understanding your unstructured data
will be required to compete
STRUCTURED UNSTRUCTURED
Reporting: Structured vs. Unstructured
STRUCTURED UNSTRUCTURED
SEARCHTask-Oriented
DISCOVERYOpportunity-Oriented
Data Retrieval Information Retrieval
Data Mining Text Mining
Unstructured Customer Service Text DataOpportunities to gain deep insight in to your customer service business
Support Case Brief Description
Support Case Activity NotesSupport
CaseCause Notes Support
Case Resolution Notes
Field ServiceRequest Description
CustomerSurvey Comments
Fields
CustomerCommunity
Threads
Employee Engagement
Survey Comments
Knowledge Base Search Strings
Project Management Tool For PS (open text fields)
CustomerSocial Media
Feeds
Unstructured Text Mining & Analytic Techniques
Sentiment AnalysisAnalyze the opinion or tone
Topic ModelingDominant theme clusters
Term Frequency (Inverse Document Frequency) Word frequency & relative importance to the set
Named Entity RecognitionProper noun determination (using NLP)
Event ExtractionRelationship between words to determine events
Similarity MatchingMatching based on similarity algorithms
Clustering“Unsupervised Learning”
Classification“Supervised Learning”
Tip #1 for Working with Unstructured Data
#TIP1#
TIP2#
TIP3#
TIP4
Avoid creating Structured Data as a BAND-AIDto get to your Unstructured Data
Tip #2 for Working with Unstructured Data
#TIP1#
TIP2#
TIP3#
TIP4
If possible, put GUIDELINES in place to Better Manage Unstructured Data Entry and Maintenance
Industry Best Practice Guidelines
Guidelines that add process steps to your workflow (avoid)
Tip #3 for Working with Unstructured Data
#TIP1
#TIP3#
TIP4
#TIP2
Consider using OPEN SOURCESOLUTIONS as a Starting Point
Tip #4 for Working with Unstructured Data
#TIP1
#TIP4
#TIP2#
TIP3COMBINE Structured and Unstructured Reporting for Enhanced Business Insight
What CUSTOMERS
DO
What CUSTOMERS
SAY Customer
Insight
PHASE
Unstructured Data Analytics Readiness
No Templates or Guidelines forData Entry & Maintenance
Templates & Guidelines
Ready for Use
Sustained Adherence &
Enhanced Reporting
Develop Industry Standard Based Guidelines with
Minimal Workflow Overhead
EffectiveTraining &
Communication
Automated MonitoringTo Manage Adherence
Realize & Communicate
Value
What does it
Look Like?
What are the Next Steps?
PLANPre-Work
DODefinition
CHECKAdopt
ACTValue
Templates & Guidelines
Consistently Used
Use Case: Support Issue / Brief Description
$68.3 M Support Revenue$19.7 M Support Cost
71% Contribution Margin118,000 Annual Case Count
105 Internal Headcount60 External Headcount
$167 Ave Cost/ Case
• Increase self service effectiveness• Remove redundant work• Gain better insight to product quality cost
Better analyze support issue brief description to find opportunities to increase customer satisfaction and reduce cost
Project Method
Project ObjectiveCompany Overview
Phase 1: Pre-Work
Pre-Work
• Unstructured brief descriptions • No templates, guidelines, or
discipline• Non-descriptive:
(i.e. “Customer needs help”)• Some system generated
(i.e. “Customer Portal Request”)
Pre-Work State
Future State Plan
1. Identify 3 categories of customer issues:“How Do I” (Q&A) Request Fulfillment
Problems
2. Put guidelines in place for each category
3. Apply Industry Best Practice
Example Brief Description Guidelines: − - Capture “What Is” vs. “What Is Desired”
− - Free of Causes, Solutions, & Effects
− - Capture using Object/Deviation Format
− - Capture in the Customer’s Context
Problems
Service Strategies, Kepner Tregoe and The Consortium for Service Innovation are Registered Trademarks
Plan
DefinitionDO • Templates & Guidelines Ready for Use
• Effective Training and Communication
Phase 2: Definition
Brief Descriptions - Before Brief Descriptions - After
Customer #42012 called in w/complaint & is not happy.
Error: “disk is full” - what should I do?
New Portal Request
We just get blinking lights and no communication. Here are the serial number: 8102-122156
The architecture postscript auto-calculates even though the input variables are invalid.
The GC application abends with multiple queue sum errors indicating the disk is full.
A virtual basic diagnostic test does not distribute the object oriented module to the desktop.
FX control does not respond even though green I/O light blinks indicating positive server communication
AdoptCHECK • Templates & Guidelines Consistently Used
• Automated Monitoring to Manage Adherence
Phase 3: Adopt Brief Description Format Compliance Report
• Sustained Adherence• Enhanced Reporting• Realize and Communicate Value
Phase 4: Value - Brief Description Text Analytics Model
ValueACT
• Similarity Matching• Named Entity Recognition
Unstructured: Brief Description
Structured: Environment Data
Duplicate Case Reports
Product L1 Product L2 Env L1 Env L2
• Sustained Adherence• Enhanced Reporting• Realize and Communicate Value
Phase 4: Value DASHBOARD: Duplicate Case Cost Overview
ValueACT
• Sustained Adherence• Enhanced Reporting• Realize and Communicate Value
Phase 4: Value DASHBOARD: Duplicate Case Cluster Report
ValueACT
Use Case: Results
Ongoing ability to leverage brief description data in order to find opportunities to increase customer satisfaction and reduce cost
Project Results
• Early detection of reoccurring issues• Ability to remove redundancy • More insight to the cost of product
quality• Increase opportunity for more
effective self service• ~$2.5 M cost savings opportunity
Project Outcome
• Realize and Communicate ValueValueACT
Other Practical Examples & Ideas
Text & trend analysis on queries submitted to the your KB
Unstructured Text Analytics
Sentiment analysis on customer survey comments to find skill gap & training opportunities
Analysis on PS/PM open-ended fields to better understand reasons for project delays
Employee loyalty analysis based on employee engagement survey comments
Case description & resolution combined to find skill gap and training opportunities
Text & trend analysis on social media and community threads
Analysis on support case notes to better determine customer adoption rates
Unstructured text analytics to predict churn and manage loyalty drivers
QUESTIONS?
